**作者：**Sun, Peize, Yi Jiang, Rufeng Zhang, Enze Xie, Jinkun Cao, Xinting Hu, Tao Kong, Zehuan Yuan, Changhu Wang, and Ping Luo

**作者单位：**香港大学，卡耐基梅隆大学，字节跳动，同济大学

发布时间：2020

发布期刊/会议：Arxiv

出版商：

**论文全称：**TransTrack: Multiple Object Tracking with Transformer

论文地址：

TransTrack: Multiple Object Tracking with Transformer

论文代码：

https://github.com/PeizeSun/TransTrack

**地位：**第一次将Transformer引入到了MOT领域并获得了不错的效果

个人理解

• 为什么：
  • 目前占据主导地位的MOT方法是复杂的多步检测跟踪管道。
  • 直接将query-key机制从SOT迁移到MOT将导致新对象的严重丢失
• 创新点：
  • 第一次将Transformer引入到了MOT领域并获得了不错的效果
  • 提出使用 一组可学习的 query（Detect Query） 来检测新出现的目标

一、摘要

In this work, we propose TransTrack, a simple but efficient scheme to solve the multiple object tracking problems. TransTrack leverages the transformer architecture, which is an attention-based query-key mechanism. It applies object features from the previous frame as a query of the current frame and introduces a set of learned object queries to enable detecting new-coming objects. It builds up a novel joint-detection-and-tracking paradigm by accomplishing object detection and object association in a single shot, simplifying complicated multi-step settings in tracking-by-detection methods. On MOT17 and MOT20 benchmark, TransTrack achieves 74.5% and 64.5% MOTA, respectively, competitive to the state-of-the-art methods. We expect TransTrack to provide a novel perspective for multiple object tracking.

• TransTrack利用了T ransformer 架构，这是一种基于注意力的 query-key 机制
• TransTrack 应用前一帧的对象特征作为当前帧的 query ，并引入一组可学习的 query 来检测新出现的目标
• 在mo17和mo20的测试中，TransTrack的MOTA分别达到了74.5%和64.5%，与最先进的方法相媲美。